Re: [MiNT] Shutdown() discussion

[Petr Stehlik <joy@sophics.cz>]

On Fri, 2003-12-12 at 10:07, Frank Naumann wrote:
> > Actually it worked OK in TOS and as it followed good examples of say
> > XHDI it looked OK.
> 
> The problem, XHDI isn't a good example, XHDI is a bad example.

I have realized it already. I was just explaining the way we thought two
years ago.

> > So what I am going to do is to propose (in ARAnyM/EmuTOS list) a step
> > back - to deprecate the NF cookie pointers and to use the illegal
> > instructions as the only and direct method of invoking the NF.
> 
> But only because you still don't want to write a /dev/aranym for these
> features for FreeMiNT and you don't get a special syscall in freemint for
> aranym.

No, that's not true. One reason is that /dev/aranym might be an overkill
for some purposes. The other reason is that /dev/aranym would force
non-MiNT users to install TSR hacks.

BTW, when I said "the only method of invoking NF" I didn't mean to stop
others from developing /dev/natfeat (/dev/aranym is a nonsense) - what I
meant was to stop people from using the pointers in cookie because that
is the real problem for next two or three thousand years (or what was
the time horizon for the virtualized memory according to Konrad :) In
other words, I welcome /dev/natfeat implementation if there is such
need.

> No again. You just wanted todo something but you don't spent much time on
> building an architecture for that and integrating it into the
> existing system architecture.

There is no existing architecture yet - people are only started thinking
about building it (GSXB in the kernel, XDD registered functions, general
VDI with graphics card support, IKBD via XDD etc). NatFeats could have
been part of this upcoming architecture.

> You just wanted to write 20 lines of code and getting anything you might
> find useful easily pushed through the emulator.

Some things are better done easily at first, than never. Just IMHO.

Petr